backlog-execute-serial

Executes backlog task files sequentially on one shared git branch using implementer and reviewer subagents.

1|Updated Dec 18, 2009
One-click install
npx skills add https://github.com/thurn/dotfiles --skill backlog-execute-serial-thurn
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: backlog-execute-serial
Source: https://github.com/thurn/dotfiles/tree/main/.llms/skills/backlog-execute-serial
Command: npx skills add https://github.com/thurn/dotfiles --skill backlog-execute-serial-thurn

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When you have a queue of task files in /tmp/backlog/ that depend on each other, running them in parallel breaks ordering guarantees. This Skill drains the backlog strictly one task at a time on a single shared git worktree and branch, so each task builds directly on the reviewed commits of the tasks before it. ## Core Features & Use Cases - Serial implementer-reviewer pipeline: Each task runs an implementer subagent, then a fresh reviewer subagent, with a fix loop (default 3 rounds) until the task is approved. - Shared linear branch: Approved commits stay on one work branch (default backlog/serial), producing a single linear branch ready to merge to master. - Automatic rollback on failure: Failed tasks are reset back to the pre-task SHA with WIP preserved on a separate branch, keeping the shared branch clean for dependent tasks. - Use Case: You generated five ordered refactoring tasks where Task 3 depends on Task 2's output. Run this Skill to execute them in numeric order, review each one, and end with one merge-ready branch. ## Quick Start Work the backlog serially by executing each task file in /tmp/backlog/ one at a time on a single shared branch.

Frequently Asked Questions about backlog-execute-serial

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run backlog tasks sequentially with dependencies?▼

Use backlog-execute-serial to process task files in /tmp/backlog/ in numeric order on one shared branch. Each task's commits land on the branch before the next task starts, so later tasks automatically build on earlier work.

What is the difference between backlog-execute and backlog-execute-serial?▼

backlog-execute runs independent tasks in parallel across sibling worktrees with an integration lane, while backlog-execute-serial runs one task at a time on a single shared branch. Choose serial when tasks have ordering dependencies; choose parallel for independent tasks and higher throughput.

What happens when a task fails review in serial backlog execution?▼

The shared branch is reset to the pre-task SHA with git reset --hard, and any committed work-in-progress is preserved on a backlog/serial-failed/ branch. The task file stays in /tmp/backlog/ for a re-attempt, and the run continues unless STOP_ON_FAILURE is set.

Does backlog-execute-serial merge the work branch into master?▼

No. The skill stops after producing the linear work branch and printing a summary of completed and failed tasks. The user performs the merge into master manually.

How many review rounds does each backlog task get?▼

Each task gets up to MAX_ROUNDS review rounds, defaulting to 3. When the reviewer requests changes, the implementer is re-dispatched in the same worktree until the task is approved or the round limit triggers a rollback.